Advanced analysis and detection of cervical cancer using NASNetLarge

Cervical cancer detection plays an pivotal role in healthcare sector. While usage of methods like visual inspection and manual examination often leads to human inaccuracy. Proposed study makes use of NASNetLarge aiming to enhance diagnostic accuracy and efficiency to tackle these challenges. Data augmentation techniques were used for a smooth training process, of the training data, extensive data augmentation techniques were employed. Using Keras, which is a Python library based on TensorFlow 2.8.2, the study implemented transfer learning which allowed the model to take advantage of NASNetLarge model’s feature extraction capabilities. The top layers of the NASNetLarge model trained on the ImageNet dataset originally, were upgraded to layers tailored for the specific classification task. Adam optimizer was used for training the model with a learning rate of 0.0001 for 27 epochs with a batch size of 8 achieving a remarkable training accuracy of 94% demonstrating its potential for aiding in the clinical diagnosis of cervical cancer. The performance of the model was evaluated on a validation set. Overall the model resulted in an accuracy of 98.33%.